How FDA and Industry Aim to Find and Eliminate Drug Recalls' Root Causes

Rooting out recalls: FDA to inspect for common causes of recalls identified in ongoing study. Findings on subpotency and B. cepacia shared. Consultant offers process for root cause analysis.

In a year that is taking shape as the most recall intensive on record, FDA said it is working to identify and thwart the most common root causes of drug recalls during facility inspections, while an industry consultant shared his method for helping companies identify the root causes of their own recalls.

FDA has listed 1,598 human drug recalls through September of this year, far exceeding the previous year's 426, as well as the 464 seen in 2000, the most recall-intensive year in the previous decade.

Rick Friedman, director of the Division of Manufacturing and Product Quality in CDER's Office of Compliance, said that for the past several years he and the office's deputy director, Joe Famulare, have been working "on bringing research to bear in our decision making."

Speaking at the PDA/FDA conference in Washington last month, Friedman said that, as part of that initiative, the agency is working with a contractor to identify major root causes of drug recalls and is bringing those findings to its investigators for use during inspections.

"We're instituting [the recall root cause analysis] permanently into our system with a new group through a statistics background, and industry background, where we can on an ongoing basis data mine and do pattern recognition for the causes of defects in the recall data we had never mined before," Friedman said.

The analysis is drawing from multiple disparate information sources, including follow-up field reports, company investigation reports and FDA establishment inspection reports.

The first of nine papers from this work appeared in the August issue of Pharmaceutical Technology, with a second expected late this year or early next. Each paper covers a different issue area such as dissolution failures or content uniformity failures.

"We are adapting our compliance programs to look at those risks and our knowledge transfer assignments and other inspectional assignments to tell the investigator a lot of the times the failures occur because of this, be sure to look at that in the inspection," Friedman explained.

Lynn Torbeck, the statistician who heads Torbeck & Associates, the Evanston, Ill., consulting firm that FDA contracted with for its recall root cause study (¹ (Also see "Torbeck to Probe Recall Data For Insights Into Root Causes" - Pink Sheet, 1 Oct, 2007.)), told the PDA/FDA meeting Sept. 15 that he is two years into this three-year project "looking at recall root cause analysis, trying to find out why do we have drug recalls and what do we do about that."

Eight Years of Class I Recalls Analyzed

Torbeck focused his remarks on early findings from an analysis of 105 Class I recalls that occurred from December 2000 to September 2008.

A study of 170 Class II recalls for 2006 and 2007 is still ongoing. "We may add 2008 to that shortly."

Torbeck updated the conference on study findings on recalls due to Burkholderia cepacia contamination as well as subpotency. Other topics he may explore include content uniformity, dissolution and transdermals.

Of the 105 Class I recalls, 49 were Rx and 56 OTC. Similarly, 49 were initiated by the company and 56 were prodded by FDA. The ratios are identical only by chance.

Of the 56 FDA-driven recalls, 14 were Rx and 42 OTC. Of the 56 OTC recalls, 34 were for undeclared new drugs.

Class I recalls increased from 2000 to 2008, particularly for OTC drugs, as FDA continues to find undeclared new drugs.

Of the Class I problems:

· 32% were for undeclared new over the counter drugs, most of which were for erectile dysfunction;

· 26% were for microbial contamination;

· 9% were for lack of content uniformity;

· 7% were associated with the heparin issue;

· 6% were for incorrect labeling;

· 5% were for tablet issues; and

· 15% were for other reasons.

Of the situations leading to the Class I recalls, 73 percent did not involve an injury. Of the rest, 12 involved one or more deaths, five involved nonspecific complaints, two involved bacterial contamination, one involved the rejection of a donor's heart; one involved a case of meningitis, one involved reports of seven pregnancies; and 10 involved other injuries.

Of the recalls associated with fatalities, seven were related to contaminated heparin. Microbial contamination, mislabeling, a defective oxygen tank and dilution error each were associated with a single recall-related fatality. No reason was given for the 12 fatality.

Of the recalls prompted by FDA, 61% were for undeclared new drugs, mostly erectile dysfunction drugs, sold over the counter. FDA has since turned its attention to also getting undeclared weight-loss drug off the market, with similar success.

Why So Many Stability Tests Show Subpotency

Torbeck's study identified 121 recalls from Nov. 8, 2000, to Oct. 23, 2008, that resulted from subpotency.

Four fifths were for prescription drugs; the rest for over the counter products. The breakdown by class was: 4 percent Class I; 38 percent Class II; and 58 percent Class III. Only 4 percent were associated with injury.

Stability failures were associated with 75 percent of the subpotency recalls. Torbeck called attention to two statistical reasons for this:

· Stability lots are tested multiple times and "the more times you test it, the more likely it is to fail;" and

· Companies set shelf life based on the average potency loss, but check each lot individually - which would only work if each lot were average. "It's statistically impossible to do. You can't set a criteria on the average and then expect all the individual values to meet that criteria."

Torbeck noted that a PQRI committee has been working on the shelf life potency criteria "and I think we can count on them to address the issue."

He added that development and stability are often done with a minimum of studies and minimum of data, which then results in poor estimates and a lack of product and process knowledge. "So that's when, if the process is marginal to begin with, then we start putting it on stability and testing it multiple times, then we have a higher probability of failure."

He concluded that one way to avoid these types of subpotency failures is to follow the approach FDA has been advocating "of designing quality into the product at the beginning of the product lifecycle."

Why B. cepacia Deserves Close Attention

Torbeck shared study findings regarding contamination by Burkholderia cepacia, an underappreciated threat.

"B. cepacia is a clear and present danger based on what we've learned so far," he said. It poses a high risk to immune-compromised people, particularly those with cystic fibrosis or those with tuberculosis or cancer, as well as babies, the elderly and pregnant women, he added.

B. cepacia is a common pathogen that exists in dirt and potable water, even distilled water. It can eat Agent Orange, forms resistant biofilms and is difficult to prevent or eliminate, he said.

"Part of the problem is complacency," Torbeck said, "because it is a very low risk for normally healthy people." However, he stressed, "it is absolutely devastating for people with cystic fibrosis." On average, B. cepacia has been found to rob 15 to 19 years from the life of cystic fibrosis patients it infects. If B. cepacia infects a cystic fibrosis patient within a year of receiving a lung transplant, it's 50% to 100% fatal according to the literature.

In fact, he said, the Cystic Fibrosis Foundation won't let you attend its annual conference if you have a B. cepacia infection.

The organism has contaminated all types of products, including mouthwash, eyewash, nasal spray, baby wipes, surgical prep cloths, skin cream and electrolyte solution - even in the presence of antimicrobial preservatives.

"In the recalls that we looked at, the products had antimicrobial preservatives in them, and yet B. cepacia was still growing in there to the point where products would smell like sour milk."

"This is not a new topic," he said, noting that FDA has warned industry about B. cepacia as far back as 1981, "but we are finding it to be much worse than we thought by doing the recall investigations."

He cited recent findings in the technical literature supporting the view that B. cepacia is a very dangerous contaminant that the industry should worry about.

And it appears that industry is getting the message: according to FDA enforcement reports, after learning that a raw material may have been contaminated with B. cepacia, McNeil Consumer Healthcare Aug. 21 initiated a Class II recall of 8 million bottles of liquid children's and infant Tylenol products.

Sterility Proposal Debated, Withdrawn

Despite such concerns, FDA's Pharmaceutical Science and Clinical Pharmacology Advisory Committee demurred when FDA Aug. 5 asked it to concur with an agency recommendation to require surgical site preparation products to be manufactured as sterile products as a way to prevent such contamination.

The committee raised so many questions that FDA tabled its recommendation.

The issue came to the fore after an FDA inspection in the summer of 2008 at Sage Products, Cary, Ind., where the agency found that prep cloths made for decontaminating surgical sites prior to incision were contaminated with B. cepacia.

A subsequent inspection of Xttrium Laboratories Inc., Chicago, showed that the chlorhexidine gluconate disinfectant solution it was providing for the cloths was contaminated with B. cepacia.

Both companies received warning letters from FDA earlier this year (² (Also see "Contamination Issues Featured in Second Quarter Warning Letters" - Pink Sheet, 1 Aug, 2009.)).

Even though there was a sense that physicians generally assume, incorrectly, that surgical prep cloths are made aseptically, the committee questioned whether FDA had made the case that they should be required to be made that way, given the likely high cost of upgrading manufacturing facilities. Members suggested it may be possible to prevent problems simply by adhering to the cGMP requirements for non-sterile products.

David Hussong of FDA's Office of Pharmaceutical Sciences said the problem typically occurs during storage when B. cepacia, if present, can consume antiseptics, eventually achieving "a phenomenal load of potential pathogens."

Hussong noted that there are two suppliers that produce the prep cloths aseptically for sale in Australia and India, but that have not obtained authorization to market them in the U.S.

Michael Jhung of the Centers for Disease Control and Prevention told the committee that in the U.S., surgical procedures have increased significantly over the past decade, while the rate of surgical site infections per 100 procedures has remained high.

However, B. cepacia was not among the top 10 pathogens listed for surgical site infections.

Recall, Shortage and Emergency Coordinator Karen Hirshfield of FDA's CDER Office of Compliance listed nine drug recalls over the past decade that resulted from microbial contamination of antiseptics, including five Class I recalls. Three involved B. cepacia.

Typically, the problems resulted from using water of low quality, such as tap water, instead of purified water, or from inadequate processes for equipment cleaning and maintenance, water qualification and testing, or from inadequate personnel training, she said.

The Need for Speed in Root Cause Analysis

Wayne Taylor, chairman of Taylor Enterprises Inc., Libertyville, Ill., shared his approach for helping clients find the root causes of their drug and device recalls.

Taylor told the PDA/FDA meeting that he has done hundreds of root cause investigations for clients, primarily for devices, but also for many drugs.

He noted that typically, drug recalls result from chemical failures while device recalls result from mechanical failures.

You find quickly with devices that "no change is too small that it can't result in catastrophic side effects." For example, there was the new IV drip that sometimes went into free flow when a certain chemotherapy drug was in the IV bag. In the lab, they found that just as the drop was wiggling loose, instead of falling, it would wick up and then down around the chamber in what was later determined to be an unidentified failure mode.

Another case involved an IV bag with two chambers separated by a weak membrane so that when squeezed, it would break, allowing the contents of both chambers to mix. However, when squeezed, the bags were splitting open, spilling their contents onto the floor. It was later found that the melt index of resin supplied to a supplier had increased by 1 degree. The change in the resin, used as a coating agent on the sheet extrusion operation, changed the sheeting properties, causing the failures. He said this case shows that no change is too small. "You need to know what your suppliers are doing. You need to know what their suppliers are doing. Anywhere in the supply chain a change can occur like this and cause a problem."

But there was another factor: the higher melt index was still within specifications, so "the full specification range was not tested out as it should have been during the initial design verification and process validation activities," he said.

Taylor went on to give tips on how to find the root cause of a recall.

"The worst possible thing you can do is have the escalating recall," he said. "You recall one week's production, then you recall a month's production, then you recall six months of production. Why does that type of thing happen? And there are plenty of examples of it happening, and it's just that we don't know what the root cause is as we started to initiate the first recall. Later we were surprised, we took our best idea of what was causing it. Later it turned out to be something else and it has expanded impact.

"So one of the things that's part of a recall, getting to root cause as quickly as possible, is extremely important to make sure that any actions that we take are in fact effective in preventing it."

Process Structured to Give Everyone Control

Taylor said he uses a structured process to uncover the root causes of recalls. Because no one person has all the information required to ascertain the root cause, "you're getting information from engineering and materials, and suppliers and quality and so on and you have to put all the pieces together."

Bad root cause investigations focus on the solutions, are based on conjecture, fixate on the initial idea of what the problem is, look for short-term Band-Aid solutions, and proceed with everyone keeping their cards hidden, Taylor said.

It's important, he stressed, to make it a cooperative team approach. "A good process removes the politics and encourages everyone to constructively participate."

On this point, he has two rules:

1. Make sure everybody participates and everybody's idea is heard and added to the list of potential causes.

2. The only way to cross out a potential cause from the list is if someone brings hard data to the group that proves it cannot be the cause. Everybody gets to participate in the examination and analysis of the data and participate in the writing of a fact statement about the data.

"So everybody's idea is heard and everybody has a veto power," he said. "Everybody has control over the process."

He noted that there are two types of situations probed in root cause analysis. There is the sudden onset problem, the root cause of which is the thing that changed. Then there is the chronic problem that is suddenly recognized, and so the root cause is the variable or variables that require adjustment.

These situations can be confused because chronic problems look just like sudden onset problems when they're discovered. "It's important as early as possible in the problem-solving process to distinguish which one of these it really is."

An excellent approach to chronic problems is the Six Sigma breakthrough improvement process, Taylor said. He has adapted that process into what he calls a Six Sigma problem-solving process to address sudden onset problems.

Taylor's Six Sigma problem-solving approach has four steps: measure, analyze, improve and control. He focused his talk on the measuring and analyzing phases, in which companies describe the sudden onset problem and identify the root cause.

In describing the problem, teams often start with a problem statement. This should determine when the problem started as well as its magnitude. Pin­pointing the start of the problem is important to distinguish it from a chronic problem and to focus suspicions on changes that occurred at the same time.

Useful tools include a process flow diagram, initial time line, a problem statement and a problem description that includes what, where, when and the problem's extent.

A particularly useful tool is the "is/is not" diagram, Taylor said. For each topic, what, where, when and extent, you ask specific is/is not questions. For example, "what specific object has the defect?" and "what similar objects could have the defect but don't?"

"The 'is not' is very important," Taylor said, "because it helps us later to identify potential causes as to what's different about the 'is' and the 'is not.'"

While it is important to let all ideas be heard, Taylor said it also is important to separate conjecture from fact. "I treat everything in the problem statement and the 'is/is not' diagram as conjecture. … My own experience has been 20 percent of what's in that 'is/is not' diagram later turns out to be wrong."

He explained that later "we'll want to transfer those things over to a fact list with data verified and everybody getting to comment and evaluate and say now can we turn this former conjecture into a fact based on objective evidence."

Also Taylor clarified that the 20 percent of wrong conjectures include many that are misleading rather than flat wrong. He recounted the case of a particulate problem in "the big room." In that case, "you look at the data later and see there are 10 different codes of product made in the big room and only two of them have this problem. So saying the problem was with the big room is a shade off. … This particular problem-solving team was called the big room particulate team, so the title of the team was misleading."

He emphasized the importance of pinpointing the start of the problem. Also, he noted that it's important to share with the team the analysis of the data rather than the data itself. In the case of an increase in particulates, change-point analysis of particle counts shows the timing and magnitude, in this case a doubling, of the change.

Eliminate the Impossible, Embrace the Improbable

Once the timing and magnitude of the change is identified, work can begin on finding the root cause in what Taylor calls the 'Analyze Phase.'

The first step is to document all of the potential causes on a "cause form." Also, he stressed that anything can be added to this list at any time.

Next all readily available data should be collected and presented to the group. However, Taylor noted that often problem investigations "collect two to three times the amount of data that's absolutely necessary to find the root cause."

To eliminate such wasted effort, it's important to quickly establish an investigation plan "that determines what additional data needs to be gathered and collected."

Then the work begins on a facts list, which lists each verified fact that the group agrees on and is supported by hard data. Each fact gets a title, a full description and a description of supporting data.

Then the group compares causes and facts. Often there can be hundreds of causes and perhaps 50 facts. To manage this array, Taylor uses what he called a "contradiction matrix." There is a row for each cause and a column for each fact. Each intersection is filled in with an X, O or A, depending on whether the facts contradict the cause, support the cause or support the cause only under certain assumptions.

The contradictory facts should carry much greater weight than the supportive facts, Taylor said. "Xs demonstrate that something could not have caused the problem. Os are only circumstantial evidence. I don't care if they have five Os next to them, if there's one X, it's been ruled out," he said.

"Finding root cause is primarily a process of elimination," he said, quoting Sir Arthur Conan Doyle's fictitious detective, Sherlock Holmes: "Once the impossible has been eliminated, what remains, no matter how improbable, is the solution to the problem."

- Bowman Cox ([email protected])